Data-driven material identification in heterogeneous viscoelastic media

Full-field material characterization techniques are increasingly improving and new techniques are emerging, providing us altogether with unprecedented access to kinematic measurements. Extracting material properties from data has been a long quest, with the challenge lying at extricating homogeneity in inference and enabling extraction of spatially-varying mechanical properties. We propose a data-driven approach here that enables identifying heterogeneous properties of linear media, either elastic or viscoelastic. We reformulate the identification problem to that of identifying a set in a phase space of possible stresses and strains and employ a hyper-plane data clustering approach algorithm to infer the pointwise material set.

Duke Scholars

Author Hossein Salahshoor Civil and Environmental Engineering